CBT-Bench Level II
Models identify cognitive distortions and classify patients’ core beliefs at two levels of detail. The evaluator compares the selected labels with the reference labels.
Slingshot AI ranks ten models by mean weighted F1 across three classification tasks. Its comparison covers Level II of CBT-Bench.
- Made by
- Mian Zhang and colleagues
- Released
- October 2024
- Kind
- Medicine
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Subject
- Psychology, Clinical medicine
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Based on
- A version of another benchmark
- In views
- Biology and medicine
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
Claude Opus 5.5 leads with 0.63 among models released by September 2026
Best score per model from Slingshot AI’s leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6.1 Terra
- and 13 more
- 1 Claude Opus 5.5 temperature=0, max_tokens=2048 0.63
- 2
Gemini 2.5 Pro Jun 2025, temperature=0, max_tokens=2048 0.63
- 3
Gemini 4 Argon temperature=0, max_tokens=2048 0.62
- 4 GPT-5 temperature=0, max_tokens=2048 0.62
- 5
Qwen3.5-122B-A10B temperature=0, max_tokens=2048 0.61
- 6 GPT-6 Astra temperature=0, max_tokens=2048 0.61
- 7 GPT-6 Luna temperature=0, max_tokens=2048 0.60
- 8 GPT-4o Aug 2024, temperature=0, max_tokens=2048 0.60
- 9 Claude Sonnet 4.5 temperature=0, max_tokens=2048 0.60
- 10
Qwen3.5-27B temperature=0, max_tokens=2048 0.59
Source: Slingshot AI’s leaderboard
| # | Model | Score | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | 0.63 | temperature=0, max_tokens=2048 | Sep 2026 | openrouter.ai | Official leaderboard |
| 2 | | 0.63 | temperature=0, max_tokens=2048 | Jun 2025 | openrouter.ai | Official leaderboard |
| 3 | | 0.62 | temperature=0, max_tokens=2048 | Sep 2026 | cloud.google.com | Official leaderboard |
| 4 | GPT-5 | 0.62 | temperature=0, max_tokens=2048 | Aug 2025 | developers.openai.com | Official leaderboard |
| 5 | | 0.61 | temperature=0, max_tokens=2048 | Feb 2026 | huggingface.co | Official leaderboard |
| 6 | GPT-6 Astra | 0.61 | temperature=0, max_tokens=2048 | Sep 2026 | developers.openai.com | Official leaderboard |
| 7 | GPT-6 Luna | 0.60 | temperature=0, max_tokens=2048 | Sep 2026 | developers.openai.com | Official leaderboard |
| 8 | GPT-4o Aug 2024 | 0.60 | temperature=0, max_tokens=2048 | Aug 2024 | developers.openai.com | Official leaderboard |
| 9 | Claude Sonnet 4.5 | 0.60 | temperature=0, max_tokens=2048 | Sep 2025 | openrouter.ai | Official leaderboard |
| 10 | | 0.59 | temperature=0, max_tokens=2048 | Feb 2026 | openrouter.ai | Official leaderboard |
Rankings by part
We read 16 parts from Slingshot AI’s leaderboard in 3 groups: task, cognitive distortion (f1 per label) and primary core belief (f1 per label). Each ranks models on one part of the test. Gemini 2.5 Pro leads 7 of them, and 6 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Cognitive distortion
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 ↑1 0.50
- 2
Qwen3.5-122B-A10B ↑3 0.48
- 3 GPT-5 ↑1 0.48
- 4 Claude Sonnet 4.5 ↑5 0.48
- 5
Gemini 4 Argon ↓2 0.47
- 6 Claude Opus 5.5 ↓5 0.47
- 7 GPT-6 Astra ↓1 0.47
- 8 GPT-6 Luna ↓1 0.46
- 9
Qwen3.5-27B ↑1 0.45
- 10 GPT-4o Aug 2024 ↓2 0.44
Primary core belief
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 ↑1 0.82
- 2 GPT-4o Aug 2024 ↑6 0.81
- 3 Claude Opus 5.5 ↓2 0.81
- 4
Qwen3.5-27B ↑6 0.79
- 5
Qwen3.5-122B-A10B 0.79
- 6
Gemini 4 Argon ↓3 0.78
- 7 GPT-5 ↓3 0.78
- 8 GPT-6 Luna ↓1 0.77
- 9 GPT-6 Astra ↓3 0.77
- 10 Claude Sonnet 4.5 ↓1 0.74
Fine-grained core belief
10 models we track have a score here.
- 1 Claude Opus 5.5 0.62
- 2
Gemini 4 Argon ↑1 0.60
- 3 GPT-5 ↑1 0.59
- 4 Claude Sonnet 4.5 ↑5 0.59
- 5 GPT-6 Astra ↑1 0.58
- 6 GPT-6 Luna ↑1 0.57
- 7
Gemini 2.5 Pro Jun 2025 ↓5 0.57
- 8 GPT-4o Aug 2024 0.56
- 9
Qwen3.5-122B-A10B ↓4 0.56
- 10
Qwen3.5-27B 0.53
All or nothing thinking
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 ↑1 0.63
- 2
Qwen3.5-122B-A10B ↑3 0.61
- 3
Qwen3.5-27B ↑7 0.58
- 4
Gemini 4 Argon ↓1 0.51
- 5 GPT-6 Luna ↑2 0.49
- 6 Claude Opus 5.5 ↓5 0.48
- 7 GPT-6 Astra ↓1 0.46
- 8 Claude Sonnet 4.5 ↑1 0.45
- 9 GPT-5 ↓5 0.42
- 10 GPT-4o Aug 2024 ↓2 0.17
Overgeneralization
10 models we track have a score here.
- 1 GPT-5 ↑3 0.40
- 2 Claude Opus 5.5 ↓1 0.40
- 3 GPT-6 Luna ↑4 0.38
- 4 GPT-4o Aug 2024 ↑4 0.38
- 5 Claude Sonnet 4.5 ↑4 0.38
- 6 GPT-6 Astra 0.37
- 7
Qwen3.5-122B-A10B ↓2 0.36
- 8
Qwen3.5-27B ↑2 0.36
- 9
Gemini 4 Argon ↓6 0.34
- 10
Gemini 2.5 Pro Jun 2025 ↓8 0.27
Mental filter
10 models we track have a score here.
- 1 GPT-4o Aug 2024 ↑7 0.13
- 2
Gemini 2.5 Pro Jun 2025 0.12
- 3 GPT-5 ↑1 0.07
- 4
Qwen3.5-122B-A10B ↑1 0.07
- 5 Claude Sonnet 4.5 ↑4 0.07
- 6
Qwen3.5-27B ↑4 0.00
- 7 GPT-6 Astra ↓1 0.00
- 8 Claude Opus 5.5 ↓7 0.00
- 9 GPT-6 Luna ↓2 0.00
- 10
Gemini 4 Argon ↓7 0.00
Should statements
10 models we track have a score here.
- 1 GPT-5 ↑3 0.76
- 2 Claude Opus 5.5 ↓1 0.72
- 3 GPT-6 Luna ↑4 0.72
- 4
Gemini 4 Argon ↓1 0.72
- 5
Qwen3.5-27B ↑5 0.71
- 6 GPT-4o Aug 2024 ↑2 0.71
- 7 GPT-6 Astra ↓1 0.70
- 8
Qwen3.5-122B-A10B ↓3 0.68
- 9
Gemini 2.5 Pro Jun 2025 ↓7 0.67
- 10 Claude Sonnet 4.5 ↓1 0.67
Labeling
10 models we track have a score here.
- 1
Qwen3.5-27B ↑9 0.45
- 2 GPT-6 Astra ↑4 0.45
- 3 Claude Sonnet 4.5 ↑6 0.44
- 4 GPT-5 0.43
- 5 GPT-6 Luna ↑2 0.43
- 6
Qwen3.5-122B-A10B ↓1 0.43
- 7
Gemini 4 Argon ↓4 0.42
- 8 GPT-4o Aug 2024 0.41
- 9
Gemini 2.5 Pro Jun 2025 ↓7 0.41
- 10 Claude Opus 5.5 ↓9 0.38
Personalization
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 ↑1 0.62
- 2 GPT-4o Aug 2024 ↑6 0.56
- 3
Qwen3.5-122B-A10B ↑2 0.54
- 4 GPT-5 0.53
- 5 GPT-6 Astra ↑1 0.53
- 6 Claude Opus 5.5 ↓5 0.53
- 7 GPT-6 Luna 0.51
- 8
Gemini 4 Argon ↓5 0.51
- 9 Claude Sonnet 4.5 0.51
- 10
Qwen3.5-27B 0.39
Magnification
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 ↑1 0.35
- 2 Claude Opus 5.5 ↓1 0.34
- 3 GPT-5 ↑1 0.33
- 4
Gemini 4 Argon ↓1 0.32
- 5 GPT-6 Luna ↑2 0.32
- 6 GPT-4o Aug 2024 ↑2 0.31
- 7 GPT-6 Astra ↓1 0.31
- 8 Claude Sonnet 4.5 ↑1 0.30
- 9
Qwen3.5-122B-A10B ↓4 0.27
- 10
Qwen3.5-27B 0.18
Emotional reasoning
10 models we track have a score here.
- 1 GPT-4o Aug 2024 ↑7 0.48
- 2 GPT-5 ↑2 0.42
- 3 Claude Sonnet 4.5 ↑6 0.39
- 4 GPT-6 Astra ↑2 0.38
- 5
Qwen3.5-122B-A10B 0.37
- 6
Gemini 4 Argon ↓3 0.35
- 7
Gemini 2.5 Pro Jun 2025 ↓5 0.33
- 8
Qwen3.5-27B ↑2 0.32
- 9 Claude Opus 5.5 ↓8 0.29
- 10 GPT-6 Luna ↓3 0.25
Mind reading
10 models we track have a score here.
- 1 Claude Sonnet 4.5 ↑8 0.72
- 2
Gemini 4 Argon ↑1 0.71
- 3 GPT-5 ↑1 0.70
- 4 Claude Opus 5.5 ↓3 0.70
- 5
Gemini 2.5 Pro Jun 2025 ↓3 0.69
- 6 GPT-6 Astra 0.66
- 7 GPT-6 Luna 0.66
- 8 GPT-4o Aug 2024 0.65
- 9
Qwen3.5-122B-A10B ↓4 0.62
- 10
Qwen3.5-27B 0.58
Fortune telling
10 models we track have a score here.
- 1 GPT-4o Aug 2024 ↑7 0.59
- 2
Qwen3.5-27B ↑8 0.57
- 3 Claude Sonnet 4.5 ↑6 0.57
- 4 Claude Opus 5.5 ↓3 0.56
- 5 GPT-6 Luna ↑2 0.55
- 6 GPT-5 ↓2 0.55
- 7
Gemini 2.5 Pro Jun 2025 ↓5 0.54
- 8 GPT-6 Astra ↓2 0.54
- 9
Gemini 4 Argon ↓6 0.53
- 10
Qwen3.5-122B-A10B ↓5 0.52
Helpless
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 ↑1 0.88
- 2 GPT-4o Aug 2024 ↑6 0.87
- 3 Claude Opus 5.5 ↓2 0.85
- 4
Qwen3.5-122B-A10B ↑1 0.83
- 5
Qwen3.5-27B ↑5 0.83
- 6 GPT-5 ↓2 0.82
- 7 GPT-6 Luna 0.82
- 8
Gemini 4 Argon ↓5 0.82
- 9 GPT-6 Astra ↓3 0.81
- 10 Claude Sonnet 4.5 ↓1 0.73
Unlovable
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 ↑1 0.84
- 2 GPT-4o Aug 2024 ↑6 0.84
- 3 Claude Opus 5.5 ↓2 0.84
- 4
Qwen3.5-27B ↑6 0.82
- 5 Claude Sonnet 4.5 ↑4 0.82
- 6
Gemini 4 Argon ↓3 0.81
- 7
Qwen3.5-122B-A10B ↓2 0.80
- 8 GPT-5 ↓4 0.78
- 9 GPT-6 Luna ↓2 0.78
- 10 GPT-6 Astra ↓4 0.75
Worthless
10 models we track have a score here.
- 1 GPT-6 Astra ↑5 0.71
- 2 Claude Opus 5.5 ↓1 0.69
- 3 GPT-5 ↑1 0.69
- 4
Qwen3.5-122B-A10B ↑1 0.68
- 5 GPT-6 Luna ↑2 0.68
- 6
Qwen3.5-27B ↑4 0.67
- 7
Gemini 4 Argon ↓4 0.67
- 8
Gemini 2.5 Pro Jun 2025 ↓6 0.66
- 9 GPT-4o Aug 2024 ↓1 0.66
- 10 Claude Sonnet 4.5 ↓1 0.65
3 models have held number one on Slingshot AI’s leaderboard since its release in October 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Slingshot AI’s leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Slingshot AI’s leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Law, finance and medicine
- MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy.
Qwen 3.8 Max 80.4%
- HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
- HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
- FinSearchComp In use Agents search the web to fetch live financial data, look up history, and run multi-step investigations, across 635 expert-written questions. Scored by answer accuracy.
Ling-3.1-flash 81.7%
- KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. o1-preview 92.7%
- Harvey LAB-AA v1.1 In use Artificial Analysis's run of Harvey's Legal Agent Benchmark on Harvey's newer private set of 120 tasks across 24 practice areas. A deliverable with a material hallucination scores zero, and the headline is the share of judges who find every rubric criterion passed. GPT-6.2 Astra 15%
Data audit
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (10 entries).
- Pass Models scored 10 models we track.
- Pass Newest score The newest is from September 2026, for Gemini 4 Argon.
- Pass Frontier models 3 of the top 10 have a score, led by GPT-6 Astra (#4).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on Slingshot AI: 10 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.